China now hosts more elite AI researchers than the US—while America still gains talent from everywhere

China accounted for 41% of elite AI researchers’ workplaces in Carnegie China’s new talent tracker, compared with 34% for the United States. In the comparable 2022 cohort, the US led 46% to 27%. Peking University also replaced Google as the highest-ranked institution in the study, while China’s share of researchers’ undergraduate origins rose from 46% to 57%.

The United States has not stopped attracting talent. It recorded a net gain of 2,145 researchers when undergraduate origin is compared with current workplace; China recorded a net loss of 1,729. The flow between the two countries remains highly asymmetric—about 30 Chinese-origin researchers work in the US for every US-origin researcher working in China—even though that ratio has narrowed from 46 to one.

The important change is retention. Sixty-nine per cent of Chinese-origin researchers now work in China, up from 57% in 2022. That remains below the US retention rate of 89%, but exceeds Europe’s 60%. South Korea and Singapore also gained, while Europe’s direction is less certain and its leading institutions fell out of the top ranking.

The study examines authors of accepted NeurIPS papers and uses complete education-and-employment histories for 10,280 people—40% of the 2025 conference’s unique authors. That sample is not random, slightly overrepresents academia and relies on self-reported OpenReview profiles. Carnegie says the China-over-US workplace finding survives broader checks; Europe’s result is more sample-sensitive. The defensible conclusion is not that America has lost AI talent, but that its old model—educate globally, then concentrate the frontier in US labs—now coexists with a far stronger Chinese domestic research system.